A Novel Point Cloud-Driven Framework for Enhanced Multi-Views Model Reconstruction and Robotic Arc Welding Trajectory Generation

Abstract

With the rapid advancement of robotics technology, robotic welding has become essential for improving the welding efficiency of large-scale complex components, while reducing the workload of welders. However, the large size and intricate weld structures present significant challenges in obtaining comprehensive point cloud and accurate welding paths, which further hinders the development and application of robotic welding in this field. To address these problems, this article proposes a novel point cloud-driven framework for enhanced multi-views model reconstruction and robotic arc welding trajectory generation of large-scale complex components. To determine the corresponding point pairs and optimal transformation used for point cloud alignment, we introduce an improved bidirectional nearest neighbor (IBNN) algorithm combined with a Levenberg-Marquardt iterative closest point (LM-ICP) approach, which enables precise and fast stitching of multi-views point clouds. We further propose an edge intensity (EI) response algorithm for efficient extraction of weld seams feature points from the point cloud, followed by B-spline curve fitting to generate smooth and accurate welding trajectories. Additionally, the welding torch pose is estimated by integrating the weld seams region (WSR) point cloud with the welding trajectories, enabling the robot to perform autonomous welding in the correct orientation. Experimental results show that the proposed framework outperforms traditional methods in both accuracy and efficiency, with a maximum error (ME) of about 0.6 mm, a root mean square error (RMSE) of approximately 0.4 mm, and a running time of around 7 s, which has a certain industrial application value.